ITB 2023 - Faster Apps That Won’t Get Crushed - Brian Klaas.pdf

Async messaging boosts app scalability. Traditional request/response cycles have limits. Services like SQS and SNS overcome scaling issues. These tools enable responsive CFML apps. Async processing improves performance under load.

Summary: Asynchronous messaging enhances web app scalability and responsiveness, outperforming traditional request/response methods and enabling efficient handling of high traffic volumes.

1. Faster Apps That Won't
Get Crushed With
Queues And Pub/Sub
Mechanisms
INTO THE BOX 2023
BRIAN KLAAS

2. Workflows

3. Workflows

4. Workflows
Click Button
Validate
Request
Check
Inventory
Process
Payment
Find Closest
Warehouse
Create Pick
Ticket
Calculate
Ship Date
Send
Confirmation
Email
Success
View

5. Workflows
Click Button
Success
View

6. Nothing’s faster
than work you don’t have to wait for.
Workflows

7. Scale and speed
are every developer’s problem.
Workflows

8. Faster Apps That Won't
Get Crushed With
Queues And Pub/Sub
Mechanisms
INTO THE BOX 2023
BRIAN KLAAS

10. Request/Response

11. Request/Response

12. We’ve got to wait around for some ill-
defined future point.
Request/Response

13. runAsync( ) in CFML 2018+
Request/Response

14. runAsync(validateRequest)
.then(checkInventory)
.then(processPayment)
.then(findClosestWarehouse)
.then(createPickTicket)
.then(calculateShipDate);
.then(sendEmailConfirmation);
Request/Response

15. var clientView = runAsync(validateRequest)
.then(processPayment)
.then(sendEmailConfirmation);
var warehouseProcess = runAsync(findClosestWarehouse)
.then(checkInventory)
.then(createPickTicket)
.then(calculateShipDate);
.then(sendShipmentDateEmail);
Request/Response

16. ColdBox asyncManager & Futures
async().newFuture( () => orderService.getOrder() )
.then( (order) => enrichOrder( order ) )
.then( (order) => performPayment( order ) )
.then( (order) => sendConfirmation( order ) );
Request/Response

17. Error handling
Retries
Throttling
New business requirements
Request/Response

18. Linear, linked flow encourages
brittle architectures
Request/Response

19. Linear, linked flow blocks
your ability to scale
Request/Response

20. How can we do better?
Request/Response

21. Event-driven

22. Something happens.
Code responds.
Event-driven

23. Database triggers
Event-driven
SQL

24. ColdBox Interceptors
Event-driven

25. Event–driven = automatic plumbing
Event-driven

26. Event-driven = easy fan-out
Event-driven

27. How do you know it’s done?
Event-driven
?

28. Orchestration tools
Event-driven

29. Orchestration = additional complexity
Event-driven
!

30. Something simpler?
Event-driven

31. Pub/Sub +
Queues

32. SNS = Simple Notification Service
Pub/Sub + Queues

33. SNS = Pub/Sub
Pub/Sub + Queues

34. SNS = One Publisher,
Many Subscribers
Pub/Sub + Queues

35. SNS Subscribers:
https endpoints (including CFML)
Email
Phone number (SMS)
SQS queue
Lambda
Kinesis Firehose
Pinpoint
Pub/Sub + Queues

36. SNS subscribers can:
Filter on specific criteria
Retry on delivery to https endpoints
Specify DLQ on completely failed delivery
Pub/Sub + Queues

37. var messageBody = { "createdOn": now()),
"userID": arguments.userID,
"event": arguments.thisEvent,
};

var messageBodyAsJSON = serializeJSON(messageBody);

var publishRequest = CreateObject('java',
‘com.amazonaws.services.sns.model.PublishRequest’)
.withTopicARN(variables.snsARN)
.withMessage(messageBodyAsJSON);

variables.snsService.publish(publishRequest);
Pub/Sub + Queues
github.com/brianklaas/awsplaybox

38. Add AWS Java SDK .jar to:
Adobe ColdFusion: cfusion/lib
Lucee: WEB-INF/lucee/lib/
ColdBox: lib/
Pub/Sub + Queues

39. SNS is about speed
Pub/Sub + Queues

40. SNS = Fan-out
Pub/Sub + Queues

41. SQS = Simple Queue Service
Pub/Sub + Queues

42. SQS = Stack of messages
Pub/Sub + Queues

43. SQS = One publisher,
one worker
Pub/Sub + Queues

44. Pub/Sub + Queues
CFML
(ack)

45. Pub/Sub + Queues
CFML
CFML
CFML

46. SQS queues can:
Perform content-based deduplication
Retry messages on failed processing
Specify DLQ on completely failed processing
Pub/Sub + Queues

47. SQS != ordered processing
Pub/Sub + Queues

48. SQS != only-once delivery
Pub/Sub + Queues

49. SQS reqiures idempotency
Pub/Sub + Queues

50. FIFO queues for order
Pub/Sub + Queues

51. Pub/Sub + Queues
github.com/brianklaas/awsplaybox
var messageBody = { "createdOn": now()),
"userID": arguments.userID,
"event": arguments.thisEvent,
};

var messageBodyAsJSON = serializeJSON(messageBody);

var sendMessageRequest =
createObject("java","com.amazonaws.services.sqs.model.SendMessageRequest")
.withQueueUrl(variables.sqsQueueURL)
.withMessageBody(messageBody);
variables.sqsService.sendMessage(sendMessageRequest);

52. var receiveMessageRequest =
CreateObject(‘java’,'com.amazonaws.services.sqs.model.
ReceiveMessageRequest').init()
.withQueueURL(variables.queueURL).withMaxNumberOfMessages(1);
var messagesArray =
variables.sqsService.receiveMessage(receiveMessageRequest).getMessages();
if (arrayLen(messagesArray)) {
var messageToProcess = messagesArray[1];
var receiptHandle = messageToProcess.getReceiptHandle();
var messageBody = deserializeJSON(messageToProcess.getBody());
// do work with the message
var deleteMessageResult =
variables.sqsService.deleteMessage(variables.queueURL,receiptHandle);
}
Pub/Sub + Queues

53. vs

54. Message Size Limit:
SNS = 256 KiB
SQS = 256 KiB
vs

55. Message Persistence:
SNS = No
SQS = Yes
vs

56. Message Ordering:
SNS = Yes
SQS = Yes, when using FIFO queues
vs

57. Message Filtering:
SNS = Yes
SQS = Yes
vs

58. Common Pattern: SNS in front of SQS
vs

59. Use cases:
SNS = Fan-out to multiple recipients
SQS = Queuing up work by processors
vs

60. SNS + SQS allow you to scale and
parallelize work safely and durably.
vs

61. Real-World Example 1

62. Example

63. Example

64. Example

65. Example

66. Example

67. Example

68. Example

69. Example
( numRequests * avg execution time )
seconds for scheduled task

70. var remainingTime = 55000;
do {
var messageProcessStartTime = getTickCount();
// get SQS message, generate and send email, delete message
var messageProcessEndTime = getTickCount();
remainingTime -= (messageProcessEndTime - messageProcessStartTime);
} while (remainingTime gt 0);
Example

71. Example
Processing–intensive tasks
Throttle processing of a batch of items

72. Real-World Example 2

73. Example

74. Example

75. Example

76. Example

77. Example

78. Example

79. Example

80. Example

81. Kinesis Data Firehose:
Can ingest up to 500,000 records/second
Stores records for up to 7 days
Manages writing to the destination
Can be queried in real-time with Kinesis Data Analytics
Example

82. Example

83. Example

84. Amazon Athena:
Query files in S3 with SQL
Store in JSON, CSV, or Parquet formats
Ad-hoc or repeated queries
Pay based on amount of data scanned
Example

85. Example

86. Example

87. Example

88. Example

89. Example

90. So how much does this cost?
Example
40 million requests = $30/month

91. Example

92. Example

93. Safe experimentation!
Example

94. Go Do!

95. Brian Klaas
@brian_klaas
Blog: brianklaas.net
github.com/brianklaas/awsplaybox

96. THANK YOU
Thanks to our sponsors